Dictation for Cursor on a Mac

Published 27 September 2026

Most of what you type into Cursor is not code. It is the instruction to the agent, the correction when it took the wrong turn, the rule that stops it doing so again, and the commit line at the end. On a Mac that is a lot of sentences typed into an editor built for code, and sentences are quicker spoken.

Deixis lets you speak them. Hold Ctrl+Option, say what you want, let go, and punctuated text is typed wherever the caret is in Cursor: the agent prompt, the file you have open, the terminal pane underneath. Your clipboard is put back afterwards. Recognition runs inside Deixis on your Mac, so the audio of your dictation stays there, and there is no extension to install and no account to make.

It is not a Cursor feature, and that is the point. When you leave Cursor for the pull request in your browser, your mail or your team chat, the chord comes with you.

Why Deixis

What it gives you in Cursor, and what it will not do

Why Deixis in Cursor

One chord from the agent prompt to everything else. The agent prompt, a rules file, a comment above the function, the source control box, the terminal pane. Then the review page in your browser and the reply in Slack. Deixis types at the caret of whichever app has focus, so nothing about it is tied to Cursor, and nothing stops working when you switch windows.
You know where your voice goes. Cursor's release notes say its voice input records the whole clip and transcribes it with batch speech-to-text after you stop. Where that transcription happens, its pages do not say, and its privacy page is silent on voice recordings either way. Deixis leaves nothing to ask: your speech is recognised on the Mac, and the audio of a dictation is never uploaded, on any setting or any plan.
It writes into every field, not only the agent's. Cursor's release notes introduce its voice input as a way to control the agent. Deixis is just as useful in a rules file, a code comment or a commit message, and it punctuates each for what it is: a comment as one sentence, a commit subject as one line, a pull request body as sentences. It can tell them apart because the names of the app and the field go with your words when cleanup hands them to a model. The window title never does, and neither does the file you have open.
It learns your stack. The Dictionary window holds the words you use: framework names, internal services, the colleague whose name always comes out wrong. They are applied on your Mac before cleanup sees anything, so useEffect stays one word. The software development pack is on from the first launch.
Point at the problem while you describe it. Double-tap the chord for free mode and tap it again for a toolbar: take the code you have selected, or grab the part of the screen that is wrong, and keep talking. It comes out as one note, your words and what you pointed at in the order you did them, pasted where you finish and saved as Markdown.
Screenshots with arrows, and screen recordings. Ctrl+Option+S freezes the screen for a region you can mark up with arrows, boxes, text and numbered steps, with a pixelate tool for anything secret, and saves it as a PNG. Ctrl+Option+V records the screen to a video on your Mac, for the change that is easier to show than to explain.
Free, with no account. Everything that runs on your Mac costs nothing and needs no sign-in. The one paid thing on a Mac is Pro, at $49 a year, which runs cloud cleanup for you so you do not need a key of your own.

What it will not do

It will not press Enter for you. Cursor's voice input takes a submit keyword in its settings, so a spoken word can set the agent running. Deixis types your words and stops there; sending them is your key press.
It does not run on an Intel Mac or on Linux. Cursor ships for Macs on Apple Silicon and on Intel, and for Linux. Deixis needs Apple Silicon and macOS 13 or later.
It understands English only. The speech model is an English model. If you prompt in another language, Deixis is the wrong tool for it.
It cannot read the diff. Deixis is voice input, not a second agent. It knows which app and which field it is typing into and never what is in them, so the change Cursor just put in front of you is invisible to it.
Your own buttons are not on the Mac yet. On Windows, a palette button can run a script of yours on a dictated note and type its answer back into the file you are in, or post the note to a webhook. Neither is in the Mac build today, and nor are meeting transcription and translation, which run on Windows and Android.
Not for files, and not for symbols. It transcribes what it records, not a recording you hand it, and brackets and flags are still faster typed.

What you get

More than dictation, on your Mac.

Hold Ctrl+Option and speak, or press a letter while you hold it: free mode, a grab, a screen recording. All of it on your Mac, all of it free, none of it behind an account.

The Deixis panel tethered to the caret in a code editor, listening, with the Mac's Ctrl+Option chord — Dictation, where your cursor is

Hold Ctrl+Option, speak, release. Clean, punctuated text lands in whatever app has focus on your Mac, with your clipboard handed back.

A finished free-mode note with a captured region and a selection interleaved with the narration — Free mode: talk and point

Double-tap the chord and narrate for as long as you like while you grab a region or take a selection. One note, what you said interleaved with what you pointed at, pasted where you finish and saved as Markdown.

A frozen screen with a dashed selection, numbered steps, an arrow and a pixelated patch — Screen grabs, annotated

Freeze the screen at the keypress, drag a region, add arrows, boxes, text and numbered steps, and pixelate what is private. Saved as a PNG in a folder you own.

A screen recording in progress, with its timer, pause and stop controls and the microphone level — Screen recording

Record the screen, pause when you need to, and stop. The video is a file in a folder on your Mac, not in somebody else's cloud.

Field by field

The same voice, punctuated for wherever you are in Cursor

Deixis knows which app and which field your words are going into — the name of the app and the name of the control, never the window title and never what is on your screen. A prompt, a message and a commit line therefore come out differently. Here is the editor and the terminal beside it, kind by kind.

Where you are	What you say	What Deixis types
A prompt to the agentthe agent prompt in Cursor	“um, pull the retry logic out of the fetch helper into its own module and give it a test for the case where the server answers 429 twice”	Pull the retry logic out of the fetch helper into its own module, and give it a test for the case where the server answers 429 twice.The filler goes, the sentence gets a capital and a full stop. A long instruction is where speaking beats typing by the most.
An instruction in a rules filea rules file, open in the editor	“prefer shad kin components and keep use effect out of the render path”	Prefer shadcn components, and keep useEffect out of the render path.Your dictionary fixes shadcn and useEffect on this machine, before anything is sent anywhere and before any cleanup sees it.
A code commenta comment above the function you are looking at	“this runs before the migration so the column may still be null on a cold deploy”	This runs before the migration, so the column may still be null on a cold deploy.A comment is one sentence, punctuated as one sentence, and left otherwise as you said it.
A commit messagethe source control message box, or the terminal	“fix the race in the release poller it was reading the feed twice on a cold start”	Fix the race in the release poller: it was reading the feed twice on a cold start.A commit line is punctuated as a line. The same words in a chat window would come out as two sentences.
A pull request descriptionthe PR body, on the review page in your browser	“this moves the retry off the hot path so the first render doesn't wait on the network um and it adds a timeout”	This moves the retry off the hot path, so the first render doesn't wait on the network. It also adds a timeout.A long field gets sentences, because Deixis knows it is a body and not a one-line box.

Cursor facts from its docs, docs_index, changelog_2_0, changelog_3_1, its privacy policy, shortcuts, downloads and mobile, read on 27 September 2026. Everything about Deixis on this page is from its README for the current release.

Worth knowing about Cursor

Cursor has voice input of its own. Its 2.0 release notes put it plainly: “Control Agent with your voice using built-in speech-to-text conversion.” You can also define custom submit keywords in settings, so a spoken word can start the agent running.
The 3.1 release notes describe the current shape: press and hold Ctrl+M to speak, with a waveform, a timer and cancel and confirm buttons while it records. It records the full clip and transcribes it with batch speech-to-text afterwards, rather than as you talk.
Where that audio goes is a question their own pages leave open. Cursor's privacy page describes prompts and code context being sent to model providers, and Privacy Mode as the switch that stops code being used for training; it says nothing about voice recordings either way.
Voice lives in Cursor's release notes rather than its documentation: the docs index carries no voice or dictation page.
The keyboard-shortcuts help page lists no voice shortcut either, among the AI ones it does list — the sidepanel toggle, inline edit, the mode menu, cycling agent modes and models, and accepting a Tab suggestion.
Where Cursor beats Deixis on reach: it publishes Linux builds as a .deb, an RPM and an AppImage alongside macOS (Apple Silicon and Intel) and Windows, and its command-line version installs from a single shell command. Deixis has no Linux build.
Cursor's mobile app is for driving agents, not for writing: it runs on iPhone and iPad, and its own help page says it has no code editing, no terminal and no file browsing, with Android planned and no date given.

Five minutes

Setting it up

Nothing goes into Cursor: no extension, no setting, no key to paste.

Install and download the model

The download is a notarized disk image. Open it and drag Deixis to Applications. The first launch shows one Download button for the speech model, which is fetched once and checked before it is kept; after that, dictation works with the network off. Settings → Updates keeps it current.

Two permissions, granted once

The first time you dictate, macOS asks for Accessibility and microphone permission. Accessibility is how Deixis hears the chord while Cursor is in front and types into Cursor's fields. Grant both in System Settings → Privacy & Security; Cursor itself needs no new permission. The first grab or recording asks, the same way, for macOS's screen recording permission.

The chord

Hold Ctrl+Option, speak, release. Double-tap for free mode, when the prompt is long and you are thinking as you go. Hold the chord a little longer and the palette shows its letters: S for a grab, V for a recording. Cursor's own voice input has a key of its own (its release notes give Ctrl+M), so you can keep both. If Ctrl+Option clashes with something else, Settings → Hotkey captures a new chord as you press it.

Your words

Spend two minutes in the Dictionary window before your first long prompt. Words to expect biases the recogniser towards the names you use, and Fixes corrects any that still come out wrong. Of the five vocabulary packs that ship, software development is already on.

What may leave

Choose once, in Settings → Cleanup. When it helps, the default, cleans your words with rules on the Mac and calls a language model only for a transform you asked for or a long run with no punctuation. Every time sends every utterance for better prose; Never keeps dictation off the network. The model can be Deixis's, your own key's, or one running on your own Mac. Whichever it is, what it gets is text: the transcript, your vocabulary, what you last dictated into that field and the names of the app and the field. Never the audio, and never the window title.

Before you ask

Questions people search for

Does Cursor have voice input of its own?

Yes. Its 2.0 release notes added control of the agent by voice with built-in speech-to-text, and submit keywords you set in settings. The 3.1 notes describe press-and-hold Ctrl+M, a waveform and a timer, and the whole clip transcribed in one batch after you stop. Voice appears in the release notes rather than the documentation, and where the audio is transcribed is left unsaid.

Which one should I use?

Use Cursor's if you only talk to the agent and want a spoken word to send the prompt. Use Deixis if you want to know the audio never left your Mac, if you talk into fields other than the agent prompt, or if you want the same chord in your browser, your mail and your chat. They use different keys, so running both is fine.

Can I dictate into the terminal pane in Cursor?

Yes. A terminal prompt is a caret like any other. It suits the prose that ends up there, a commit message or an answer to a question a script is asking, better than a command full of flags.

Does Deixis run on an Intel Mac?

No, and not on Linux either: the Mac build is for Apple Silicon on macOS 13 or later. Cursor publishes builds for both kinds of Mac and for Linux, so it reaches machines Deixis does not. Where Deixis runs keeps the state of every build.

Is it free on a Mac?

Yes. Nothing Deixis does on the Mac is metered, and none of it asks you to sign in: dictation, free mode, the dictionary, screen grabs and screen recording. If you want a model to tidy your prose, your own provider key or a model on your Mac makes that free too, and Pro, at $49 a year, runs it for you.

Keep reading

How to type by voice in any app on a Mac, with the speech recognised on the Mac itself

How to type by voice in any app on a Mac: hold Ctrl+Option, speak, let go. Speech recognition runs on your Mac, free and with no account.

Dictation on a Mac, compared with Deixis

MacWhisper, Superwhisper, VoiceInk, Wispr Flow and Apple Dictation, set beside Deixis on a Mac: where speech is recognised, what leaves, the price.

Dictation for VS Code on a Mac

Hold Ctrl+Option to dictate into VS Code chat, comments and the terminal, then into every other app on your Mac. Recognised on the Mac, no sign-in.

Dictation for Terminal and iTerm2 on a Mac

Hold Ctrl+Option and speak into Terminal or iTerm2 on your Mac. Agent prompts, commit messages and heredocs, recognised on the Mac, with no account.

Voice dictation for Obsidian on a Mac

Hold Ctrl+Option and talk into any Obsidian note on your Mac. No plugin, speech recognised on the Mac itself, and Markdown notes you own. Free.

Dictation for Slack on a Mac

Hold Ctrl+Option and speak into Slack on your Mac: replies, threads, DMs and canvases. Speech recognised on the Mac, with no account and no admin.

Two keycaps, Ctrl and Win, held down and lit teal from underneath.

Hold a key. Speak. Keep working.

Deixis is free, needs no account, and runs on your own device. Everything on this page is in the current release.

Download Deixis for Windows or Mac Every byte that leaves

v1.2.2 · 64‑bit Windows 10/11 · macOS 13 on Apple silicon · 51 MB · nothing else to install

Facts about Cursor checked on 2026-09-27 against cursor.com. Deixis facts are from its README for the current release. Something out of date? Tell us.